deepseek R1

app.py

@@ -1,70 +1,41 @@
-import streamlit as st
 import os
+import streamlit as st
+from huggingface_hub import InferenceClient
 from dotenv import load_dotenv
-from transformers import (
-    pipeline,
-    AutoConfig,
-    AutoTokenizer,
-    AutoModelForCausalLM
-)
 
-# Load environment variables from .env (if you’re using one)
+# 1. Load .env (if using)
 load_dotenv()
 
+# 2. Instantiate the InferenceClient once
+client = InferenceClient(
+    provider="auto",
+    api_key=os.environ["HUGGINGFACEHUB_API_TOKEN"]
+)
+
 st.set_page_config(page_title="Educational Chatbot")
 st.title("🎓 Educational Chatbot")
 
-@st.cache_resource(show_spinner=False)
-def load_model():
-    # 1. Load the remote config (with trust_remote_code)
-    config = AutoConfig.from_pretrained(
-        "deepseek-ai/DeepSeek-R1",
-        trust_remote_code=True
-    )
-    if hasattr(config, "quantization_config"):
-        config.quantization_config = {}     # ← use {} instead of None
-
-    tokenizer = AutoTokenizer.from_pretrained(
-        "deepseek-ai/DeepSeek-R1",
-        trust_remote_code=True
-    )
-    model = AutoModelForCausalLM.from_pretrained(
-        "deepseek-ai/DeepSeek-R1",
-        trust_remote_code=True,
-        config=config
-    )
-
-
-    # 4. Build the text-generation pipeline
-    gen = pipeline(
-        "text-generation",
-        model=model,
-        tokenizer=tokenizer,
-        trust_remote_code=True,
-        device_map="auto"        # or remove for CPU-only
-    )
-    return gen
-
-# Load the model once
-generator = load_model()
-
-# Initialize chat history
 if "history" not in st.session_state:
     st.session_state.history = []
 
-# User input box
+# 3. Chat input
 user_input = st.text_input("Ask me anything:")
 
-# When user enters a question
 if user_input:
     try:
-        outputs = generator(user_input, return_full_text=False)
-        reply = outputs[0]["generated_text"].strip()
-        st.session_state.history.append(("You", user_input))
-        st.session_state.history.append(("Bot", reply))
+        # 4. Make the chat-completion API call
+        completion = client.chat.completions.create(
+            model="deepseek-ai/DeepSeek-R1",
+            messages=[{"role": "user", "content": user_input}],
+        )
+        reply = completion.choices[0].message["content"]
     except Exception as e:
-        st.session_state.history.append(("Bot", f"⚠️ Error: {e}"))
+        reply = f"❌ API Error: {e}"
+
+    # 5. Append to session history
+    st.session_state.history.append(("You", user_input))
+    st.session_state.history.append(("Bot", reply))
 
-# Display chat history
+# 6. Render chat history
 for sender, msg in reversed(st.session_state.history):
     st.markdown(f"**{sender}:** {msg}")

requirements.txt

@@ -3,3 +3,5 @@ python-dotenv
 transformers>=4.40.0
 torch>=2.1.0
 accelerate
+huggingface-hub>=0.14.1
+